Largest open experiment: frontier models close 82% gap to human record in autonomous AI research
inductionheads · x · 2026-08-17
PrimeIntellect ran the largest open experiment on how frontier models do AI research. 100+ autonomous runs across 10+ models, sandboxed on 8xH200s for up to 8 days, iterating on the nanoGPT optimizer track. Best runs closed 82% of the gap to a record built by dozens of humans over months. FakePsyho comments that optimization problems are better at testing creativity and thus better proxies for RSI than ML problems.
Related event: Results of Largest Autonomous AI Research Experiment Open Sourced(13 posts)→
More from AGI Musings
- Math breakthroughs may have limited impact due to human digestion bottleneck — rbhar90 · 2026-08-17
- US per capita power consumption peaked at dot-com,暗示 scaling limits — jwt0625 · 2026-08-17
- Tech Industry Criticized for Misusing 'Singularity' Term — MatthewMcAteer0 · 2026-08-17
- Monitoring Is Not a Panacea for AI Safety, Alignment Is Key — tszzl · 2026-08-17
- AI code generation will create massive digital SKUs, transforming the payments market — arampell · 2026-08-17
- Byte magazine archives: Comparing early microcomputer evolution with AI's speed — PTrubey · 2026-08-17